2020 IEEE International Symposium on Circuits and Systems (ISCAS)

1 / 2019-10-15 14:10:28

Improved Vehicle Tracking Method Based on MDNet

intelligent transportation; MDNet; target tracking; instance segmentation

In the field of intelligent transportation, many factors can affect the tracking results of moving vehicles in the video, such as the background complexity, the illumination change, occlusion, scale transformation and so on. In order to solve the drift problem and improve the tracking accuracy in the vehicle target tracking process, this paper proposes an improved vehicle target tracking algorithm based on MDNet. By combining the instance segmentation method with the MDNet algorithm, the background and vehicle targets can be distinguished remarkably, which improves the accuracy of vehicle target tracking. Experimental results on public datasets show that this improved method is superior to most trackers.